Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Decision Trees Recurrently Through Communication
Stephan Alaniz, Diego Marcos, Bernt Schiele, Zeynep Akata
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use recurrent hard binary attribute decisions instead of a classical decision tree | RDTC · AWA2 | mean per-class accuracy: 92.3% vs 98.0% | Improved |
| use recurrent hard binary attribute decisions instead of a classical decision tree | RDTC · CUB | mean per-class accuracy: 43.5% vs 78.1% | Improved |
| use recurrent hard binary decisions instead of dNDF | RDTC · CUB | mean per-class accuracy: 73.8% vs 78.1% | Improved |
| add human-annotated attribute loss | RDTC · CUB | mean per-class accuracy: 78.1 ± 0.2 -> 77.9 ± 0.6 | Hurt |
| add human-annotated attribute loss | RDTC · AWA2 | mean per-class accuracy: 98.0 ± 0.1 -> 98.1 ± 0.0 | Mixed |
| combine explicit memory with LSTM tree-state encoding (ablation) | RDTC · CUB | mean per-class accuracy: 73.0 -> 78.1 | Improved |
| combine explicit memory with LSTM tree-state encoding (ablation) | aRDTC · CUB | mean per-class accuracy: 69.1 -> 77.9 | Improved |
| encode tree state with explicit memory rather than only the LSTM (ablation) | aRDTC · CUB | mean per-class accuracy: 69.1% vs. 76.4% | Improved |
| apply classification loss at every communication step rather than only at the leaf (ablation) | RDTC · CUB and AWA2 | intermediate-step accuracy | Improved |
| apply classification loss at every communication step rather than only at the leaf (ablation) | RDTC · CUB and AWA2 | final accuracy | No clear effect |
| apply classification loss at every communication step rather than only at the leaf (ablation) | RDTC · CUB and AWA2 | number of tree nodes | Improved |
| increase attribute-loss weight λ toward 1.0 (ablation) | aRDTC · AWA2 and CUB | classification accuracy | Hurt |
Closest papers by meaning
- Training Decision Trees as Replacement for Convolution Layers (AAAI 2020 · similarity 0.88)
- Generative Model for Decision Trees (AAAI 2024 · similarity 0.84)
- Multistage Defer Trees for Hybrid Interpretability: If at First You Can't Succeed, Tree Again (NeurIPS 2026 · similarity 0.84)
- ViTree: Single-Path Neural Tree for Step-Wise Interpretable Fine-Grained Visual Categorization (AAAI 2024 · similarity 0.84)
- Near-Optimal Decision Trees in a SPLIT Second (ICML 2025 · similarity 0.84)
- Branches: Efficiently Seeking Optimal Sparse Decision Trees via AO* (ICML 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
